Bottom line first. The family maxim holds that half of all advertising is wasted, and no one knows which half. Amazon's next chapter will be judged by a sharper version of that problem. The transaction volumes that built the company's reputation are no longer where the story lives. Global e-commerce sales are forecast at $4.9 trillion in 2026 9, but the doors have largely been built: online penetration is projected to rise only from 24% in 2025 to 27% in 2031 9, and annual online migration is forecast at roughly $80 billion through 2031 — far below the $240 billion recorded in 2020 9.
When floor traffic stops compounding, a merchant stops counting door swings and starts counting revenue per customer. That is where Amazon now competes. The sector's incremental economics are shifting toward advertising, subscriptions, logistics and payments rather than the retail transaction itself 9. The investment case is therefore about mix, ecosystem control and execution — and the question that should follow every number below is the same one that follows every invoice: how do you know it works?
The Advertising Counter: Bankable Margins, Unverified Books
Start with the numbers that favor the platform. Advertising at large mature platforms is described as carrying 60–80% EBIT margins 9, with additional monetization available from fulfillment and other merchant services. Amazon held about 40% of global retail-media share in 2026, up from roughly 30% in 2021 9, and Sponsored Products remains its largest advertising offering 1,13.
These are excellent margins for selling shelf space beside your own shelf. They also describe a closed loop: one company controls the catalog, the search results, the ad inventory and the sales data. The question is not whether retail media works, but how an advertiser knows it works — which campaigns produced purchases that would not have happened anyway. The material does not settle that incrementality question, and honest analysis should not pretend otherwise. What it does establish is that this revenue mix raises the stakes on two items: advertiser trust and regulatory resilience. Both grow more valuable as the mix tilts this way.
Discovery: Percentages Are Not Denominators
Now the channel everyone wants to discuss. AI-referred traffic to U.S. retail sites rose 693% year over year during the 2025 holiday season 6, and Adobe Analytics found that visitors arriving through AI channels converted at rates 60% above other traffic sources 14. Impressive percentages. Every direct-mail operator has seen impressive percentages on small lists.
Check the base. Generative-AI traffic represented only 0.9% of MercadoLibre visits in July 2026 9, despite growing roughly 24-fold from mid-2024 9. Growth rates without denominators are how waste hides. The honest reading: the channel is real, early, and not yet a replacement for anything.
The Amazon–YouTube partnership is more concrete. Creators can tag Amazon products directly in videos and livestreams, earning commissions on resulting purchases 5, and product tags generate about twice the engagement of affiliate links in video descriptions 2. That is lower friction at the point of intent — the catalog insert, relocated to where attention already sits. Together the evidence supports a modest claim, not a dramatic one: discovery is becoming more distributed, and Amazon benefits by being present at high-intent entry points rather than by winning every search.
India: Strong Growth, Itemized Costs
Indian e-commerce GMV is estimated at about $100 billion in 2026 and forecast to grow at an 18% CAGR to $141 billion in 2028 9. A merchant cannot ignore that. The complication is quick commerce, which is growing twice as fast as the overall market and is forecast to account for roughly one-third of Indian e-commerce by fiscal 2030 9. Its expansion beyond grocery into fashion, electronics, beauty and general merchandise brings it into more direct competition with Amazon India and Flipkart 7, threatening conventional marketplace and parcel-logistics volumes.
Two pressures follow. Quick commerce creates a strategic need for localized inventory and fulfillment capabilities. And Amazon, Flipkart and Delhivery's parallel seller-fee increases point to concentrated logistics and marketplace pricing power 8, which tightens margins for merchants even as the platforms protect their own economics. That tension gives sellers reasons to diversify channels and inventory. For Amazon in India, seller retention and logistics integration are not support functions. They are the competitive variables.
China: The Subsidy Ledger
China shows what ecosystem expansion costs when it is bought with subsidies. Alibaba and JD.com incurred combined food-delivery losses of RMB 132 billion between the second quarter of 2025 and the first quarter of 2026 9, and 2026 food-delivery EBIT is forecast at negative RMB 33 billion for Alibaba and negative RMB 25 billion for JD. JD's commerce GMV is forecast to grow only 3% in 2026, versus 19% for Douyin 9 — slower growth, absorbed alongside margin volatility.
The Street has not entirely soured. Goldman Sachs rated JD.com Buy with a $43 price target implying 47% upside 9; the same material identifies sharper competition, GMV-slowdown risk and margin swings 9. Both can be true at once. The comparative lesson for Amazon concerns discipline rather than direction: adjacent-service expansion is attractive when it reinforces a profitable flywheel, and dilutive when it becomes a subsidy-led battle that erodes the very returns ecosystem monetization is meant to deliver.
Regulation: When the Cash Register and the Referee Share a Desk
Large retail-media platforms control data, inventory, technology and pricing rules inside closed ecosystems 3,12, an arrangement that has drawn scrutiny over conflicts between the marketplace and advertising roles. Amazon's products at issue include Sponsored Products, Sponsored Brands and Display ads alongside search results 11.
The Google ad-tech outcome is the relevant precedent. A U.S. court did not require Google to sell its ad exchange, but imposed behavioral restrictions on use of its integrated ad-tech stack 4,10. This does not establish a direct outcome for Amazon, and I will not pretend it does. It raises the strategic premium on auction transparency, clear seller economics and demonstrable advertiser value — the three things a platform would need to produce if asked. A merchant who can prove his prices are fair sleeps better than one who must argue they are.
What a Careful Merchandiser Would Measure
Four conclusions, each stated as a measurement discipline rather than a celebration:
- Mix over migration. Amazon's principal opportunity is converting mature e-commerce scale into higher-margin ecosystem revenue — retail media and fulfillment chief among them — rather than waiting for penetration growth that is no longer coming. The number to watch is incremental margin per customer and per merchant, not transaction volume.
- Hedged discovery. AI and creator-led shopping are better read as a redistribution of discovery than as a proven replacement for marketplaces. The YouTube integration is a practical hedge, and its value should be judged by commissions and conversion, not by engagement percentages alone.
- India requires operations, not optimism. Growth toward $141 billion by 2028 is real, but quick-commerce competition and merchant-margin pressure make logistics capability and seller retention the decisive variables.
- Transparency as a durable asset. The durability of the integrated model will depend not only on scale, but on whether Amazon's advertising and marketplace practices withstand rising expectations for transparency and non-preferential conduct.
The history of advertising is a history of unmeasured waste. Amazon has built the most measurable retail machine in commerce. The open question — the one that should govern how you read every number above — is whether it will hold its own ecosystem to the same standard of proof it sells to everyone else.